What shipped lately.

Test any agent against last week’s real work before it goes live. Replay shows every step the agent would have taken and flags anything that would have needed approval.

• Replay up to 30 days of history

• Side-by-side diff against what your team actually did

• Export replay results as CSV

Set a monthly spend cap and a per-action ceiling for every agent. When an agent hits a limit it pauses and asks the owner in Slack.

Overnight runs now process queues in parallel, so large backlogs finish up to 3x faster. Morning summaries arrive before 7 am in every time zone.

Agents can now escalate straight to a Slack channel or person, with the full context attached and one-click approve or reject buttons.

Fixed an issue where audit log exports over 10,000 rows could time out. Exports are now streamed and include tool call details.